AI often has a hard time learning from rare events simply because there is not enough data. To help solve this, we have developed a technology that can generate large amounts of realistic network data from just a small sample.
At this booth, you will be able to visually explore the kinds of data our system creates-such as network packets, system logs, and alarms.
At this exhibition, we will present a new technology that improves both the efficiency and sophistication of procedure creation for network (NW) construction operations. In collaboration with NTT-ME, we will showcase our work on automatically creating new procedures by leveraging existing ones.
By using the LLM-based planner ATOMN, which is currently being developed at our research laboratories, the system can autonomously and accurately create procedures while continuously observing the current system state and the verification environment. This capability significantly reduces engineers' operational burden and is expected to improve productivity and stabilize quality in network construction operations.
Lessons learned from recent major outages have clarified that carrier networks need to be more robust, thus it is more important than ever to rapidly isolate and resolve faults. While the handling of simple failures has been automated using rule-based procedures, that of complex failures remains difficult to automate and requires manual intervention by operators, leading to prolonged recovery times. In this exhibit, we will introduce our efforts toward the commercial deployment of a self-evolving Zero-Touch Operation Framework designed to accelerate failure handling using network AI. Using this approach, we aim to accelerate the resolution of complex failures.
At this exhibition, we will introduce an initiative to significantly improve the efficiency of maintenance and operations for communication networks, which are becoming increasingly complex year by year.
We are attempting to integrate an AI agent currently being developed by NTT Field Technology with Konan, a fault localization system being developed by the Access Network Service Systems Laboratories.
This integration leverages the Model Context Protocol (MCP), which enables different AI systems to exchange information using a common set of rules. By adopting MCP, multiple AI systems can smoothly collaborate while sharing roles and responsibilities.
Through these technologies, we aim to move beyond conventional automation that simply follows predefined procedures, toward the realization of autonomous operations in which AI agents assess situations on their own and independently analyze and respond to faults.
This exhibit will showcase our efforts to enhance network operation and maintenance by integrating the NW information infrastructure NOIM and generative AI. NOIM centrally manages network configurations using a general-purpose data model, providing a platform where AI can correctly understand configuration, connections, and dependencies. It minimizes the need for custom AI tuning, enabling AI agents to be rapidly deployed and achieving high-precision inference and response reproducibility using a local LLM even in closed environments handling sensitive information. This technology accelerates the adoption of AI in network operations.
This exhibition will introduce two technologies that automate PC‑based operations by automatically generating and adapting processes.
The first technology constructs operational processes by combining procedures extracted from operation logs and business documents with decision rules derived from business documents, while reflecting the judgment patterns of experienced operators.
The second technology estimates the impact of changes in operations or rules and reconfigures existing processes accordingly, enabling processes that adapt to evolving situations.
Combined with RPA, DAP, and related solutions, these technologies advance the automation of non‑standardized operations.
In this exhibition, we will present a technology designed to enhance the stability of wireless communication quality at event venues by forecasting visitors' anticipated network usage and accordingly optimizing wireless network control.
This technology integrates external data sources, including meteorological conditions and ticket sales information, with real-time wireless network metrics to highly accurately predict network demand. On the basis of these forecasts, the system proactively adjusts wireless network parameters to ensure that capacity and resources are appropriately allocated in advance.